Research on Automatic Tracking Method of Marker Points in Sports Image Sequence Based on Feature Matching

Abstract

Generally speaking, the methods of automatic tracking and recognition in sports image sequence analysis can be divided into two categories: first, template matching method, which compares each template image with all sub-images in the search area, finds out the most similar sub-image and makes the sub-image a new template, and repeats the above process in the corresponding search area of the next adjacent image; second, feature matching method, which compares the features of the sub-images in the search area and finds the seal. Sports image collection is dynamic, imaging is vague, and the distribution structure of dynamic feature marks is complex. Automatic tracking of dynamic feature marks in sports scenes is the key to realize moving image recognition. In this paper, feature matching is introduced into the automatic tracking method of mark points in sports image sequence. The basic condition that the image to be registered is collinear with the corresponding line segment on the reference image is used to establish the image deformation model, and the full automation of sequence image registration process is realized through the automatic extraction and automatic matching of line segment features.